Post-Rejection Follow-up Sampling: A Methodology for Counterfactual Outcome Measurement in Algorithmic DEX Trading
This paper introduces Post-Rejection Follow-up Sampling (PRFS), a methodology that tracks the real-time market performance of tokens rejected by algorithmic DEX trading systems to generate a dataset of 67,000 forward-outcome observations, thereby enabling the evaluation of filter precision against actual market results rather than synthetic backtests.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a talent scout for a massive, chaotic music festival. Every day, your job is to scan thousands of new bands trying to get on stage. You have a strict set of rules (filters) to decide who gets in.
The Problem: The "Black Box" of Rejection
Usually, when your system rejects a band, that's the end of the story. You see the bands that did play and how the crowd reacted. But you never know what would have happened to the bands you turned away. Did they have a secret hit song? Were they terrible? Did they vanish from the face of the earth?
In the world of automated trading on Decentralized Exchanges (DEXs), this is a huge blind spot. The computer algorithms reject thousands of potential tokens (digital assets) every day. Traditional methods try to guess what would have happened to these rejected tokens by looking at old price charts, but those charts are like looking at a map of a city that has already changed. They don't tell you what actually happened to the specific tokens your system said "no" to.
The Solution: The "Post-Rejection Follow-up" (PRFS)
This paper introduces a clever new method called Post-Rejection Follow-up Sampling (PRFS).
Think of PRFS as a dedicated "stalker" team (in a very polite, data-gathering way) that follows the bands you rejected.
- The Decision: Your main trading system says "No" to a token and logs the reason.
- The Follow-up: Instead of ignoring that token, the PRFS system starts watching it closely. It checks the token's price and how much money is available to trade it (liquidity) at regular intervals.
- The Timeline: It keeps watching for up to 24 hours.
- First Hour: It checks very frequently (like checking your phone every 5 minutes) because that's when things usually change fast.
- Later: It checks less often (like once an hour) as time goes on.
- The Stop: It stops watching if the 24 hours are up, if the token disappears from the market (like a band quitting before the show), or if the system hits a technical limit.
What This Achieves
By doing this, the researchers can finally see the "counterfactual" outcome. They can answer: "If we had actually let this rejected token in, would it have made money, or would it have been a scam?"
This isn't about guessing or simulating; it's about measuring reality. It turns the "what ifs" into "what actually happened."
The Experiment
The researchers tested this over eight days in April 2026 on the Solana blockchain.
- They tracked 2,997 times their system rejected a token.
- They managed to get at least one update on about 55% of those rejections.
- In total, they collected 67,000 data points (price checks) on these rejected tokens.
The Results (The "Bimodal" Surprise)
The data showed a very interesting split, like two different types of rejected bands:
- The "Ghost" Bands: About half the rejected tokens disappeared almost immediately (within the first check). In trading terms, these were likely scams or tokens that ran out of money instantly. The system successfully rejected them, and the follow-up confirmed they were gone.
- The "Long-Lived" Bands: The other group stayed around for the full 24 hours. Some of these were actually good tokens that the system rejected by mistake, while others were just boring tokens that didn't move.
Why This Matters
The paper argues that this method is a better way to tune trading algorithms. Instead of guessing, you can now look at the actual fate of the tokens you rejected and say, "Hey, our filter is too strict on this type of token," or "Great, our filter caught that scam."
Limitations
The authors are honest about the constraints:
- The Window: They only watched for eight days. Market conditions change, so this is just a snapshot.
- The Density: For some tokens, they didn't get enough data points to draw a perfect line of how the price moved. They got a few dots, but not a smooth curve. This is because the data source (the price oracle) has limits on how many requests it can handle.
- The "Disappearance" Mystery: If a token vanishes right after rejection, the system assumes it was a scam. The paper admits this is a good guess, but they haven't proven it 100% against independent records yet.
In Summary
This paper doesn't claim to have found a magic money-making formula. Instead, it offers a new tool for measurement. It's like giving a chef a way to taste the ingredients they threw away, so they can learn exactly which ones were bad and which ones they accidentally tossed out. This helps them cook better meals (make better trading decisions) in the future.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.